Nate Notebook

install.packages("useful")

Installing package into 㤼㸱C:/Users/nathd/Documents/R/win-library/4.0㤼㸲
(as 㤼㸱lib㤼㸲 is unspecified)
trying URL 'https://cran.rstudio.com/bin/windows/contrib/4.0/useful_1.2.6.zip'
Content type 'application/zip' length 162666 bytes (158 KB)
downloaded 158 KB

package ‘useful’ successfully unpacked and MD5 sums checked

The downloaded binary packages are in
    C:\Users\nathd\AppData\Local\Temp\RtmpY3R07U\downloaded_packages

Get nflfastR data

pbp_2017 <- nflfastR::load_pbp(seasons = 2017)

pbp_2017 %>% head(15)

Get Big Data Bowl Data

#play by play data (looks like it is just a few weeks of data in the 2017 season)
plays <- vroom("https://raw.githubusercontent.com/nfl-football-ops/Big-Data-Bowl/master/Data/plays.csv")

Rows: 14,193
Columns: 27
Delimiter: ","
chr  [ 8]: possessionTeam, yardlineSide, offenseFormation, personnel.offense, personnel.defense, SpecialTeamsPlayType, PassResult, playDescription
dbl  [16]: gameId, playId, quarter, down, yardsToGo, yardlineNumber, defendersInTheBox, numberOfPassRushers, HomeScoreBeforePlay, VisitorScoreBeforeP...
lgl  [ 2]: isPenalty, isSTPlay
time [ 1]: GameClock

Use `spec()` to retrieve the guessed column specification
Pass a specification to the `col_types` argument to quiet this message

#tracking data for gameid 2017090700 -- may not use this at all
#tracking <- vroom("https://raw.githubusercontent.com/nfl-football-ops/Big-Data-Bowl/master/Data/tracking_gameId_2017090700.csv")

plays %>% head(5)

Clean plays dataframe

#SPLIT PERSONNEL AND ADD TO DATAFRAME

#split string for the defensive personnel
split_def <- stringr::str_split_fixed(plays$personnel.defense, " ", n = 7) %>% 
                          data.frame() %>% 
                          select(X1, X3, X5)
names(split_def) <- c("DL", "LB", "DB")

#split string for the offensive personnel
split_off <- stringr::str_split_fixed(plays$personnel.offense, " ", n = 7) %>% 
                          data.frame() %>% 
                          select(X1, X3, X5)
names(split_off) <- c("RB", "TE", "WR")

#add to plays data and create plays_pers (plays personnel)
plays_pers <- plays %>% cbind(split_def) %>% cbind(split_off)

#view first 20 rows
plays_pers %>% head(20)

Bind Big Data Bowl and nflfastR data

Filter out unwanted plays

Possible Classifications:

• Pass
  • short/deep (pass_length)
  • air_yards (possibly bin this off to get more possibilities than above)
  • pass_location
• Run
  • middle/right/left (run_location)
  • run_gap (combine with above variable – maybe just use this one with middle because the side may not really matter but the gap does)

Simplify classifications to one variable

Features to use in model:

• defensive personnel (defendersInTheBox, DL, LB, DB)
• game situation (quarter, down.x, yardToGo, yardline_100)
• numberOfPassRushers

Get model data

write.csv(model_data, "Model Data?.csv")

cannot open file 'Model Data?.csv': Invalid argumentError in file(file, ifelse(append, "a", "w")) : 
  cannot open the connection

Run Models

POSSIBLE ADD INS

• offensive formation
• yardlineSide
• rolling epa/play on passes and rushes for offense and defense
• qbr/defensive rating
• more in depth game information (time of the clock, etc.)
• weather